October 2006 Volume 9 Number 4

Agent’s Mind
Local
Domain
KB
XAPML-SEML
tables
User Input: Communicative Goal
SEML
XAPML
MIND
Body Animation
Generator
SEML
Wrapper
Figure 2. Mind-Body Architecture
Student profile
The task of this module is to generate the specification of the Agent’s move at the meaning level. Planning is a
task hard to compute, then it is not feasible to have a planner running on an handheld computer, moreover
MyCoach has to run on and off-line. For this reason, we adopted a schema based approach.
Given a communicative goal, derived by an interpretation of the user input, the Agent’s Mind selects from its
plan library the schema or a combination of schemas that allow to satisfy that goal. The result is an XAPML
(eXtended APML) specification of meanings to be communicated.
APML is a markup language designed to represent the communicative functions (De Carolis et al., 2003). Each
tag of this language is used to add some affective accent to the content of a sentence. For instance the
tag is used to specify what is the communicative speech act of the sentence (Pelachaud and
Poggi, 1998).
The extension of APML includes the following information that were not present in the first version: i) besides
the specification of what the agent has to communicate it is possible to specify the information background
(access to more details, visualization of relevant domain objects, etc.); ii) the focus attribute of the
tag allows to establish which is the main topic under discussion and to update the student profile
with information about what has been communicated to the student; iii) the voice attribute has been added since,
in case the student interact through its handheld device, the voice output has to be generated on the server side
and then passed to the client for being played together with the body animation. These recipes have been
designed on the basis of the results of a survey, in which the students were asked to rate a set of information
about the department and student life. Figure 3 shows results of second year students.
Analysing the survey results, major interests seem to concern information about modalities for choosing optional
exams for completing the third year curricula, accessing student facilities, setting an examination test, get credits
and recover from formative debits. Therefore, we developed XAPML schemas for answering to these families of
questions and for getting more detailed answers once selected a topic question. These schema allows to organize
the sequencing of the answer and to manage the multimodality of the interaction, combining text, voices, and
expressions of the face.
31